Misreporting of energy intake estimated by food records among adolescents in Lao People’s Democratic Republic: a cross-sectional study

Abstract

Background

Accurate dietary assessment data is vital for understanding dietary patterns and the associated nutrition and health outcomes. However, research on misreporting from self-reported dietary assessment tools has remained limited in low- and middle-income countries. Our study aims to investigate energy intake misreporting and its determinants among adolescents in Lao People’s Democratic Republic (PDR).

Methods

We conducted a cross-sectional secondary analysis of seven-day estimated food records collected from 405 adolescents aged 10–18 years attending secondary schools in Vientiane Province, Lao PDR, between March and April 2019. We identified energy intake misreporting using the Goldberg cut-off method, and analyzed factors associated with energy intake underreporting using multivariable logistic regression analyses. We compared the percentage energy intake from each food group between plausible reporters and underreporters, using the Mann-Whitney U-test.

Results

Approximately 61.2% of participants underreported and 0.3% overreported their energy intake. Older adolescents (15–18 years) were 50% less likely to be classified as underreporters than younger adolescents (10–14 years) (Adjusted Odds Ratio [AOR] = 0.54, 95% CI: 0.34–0.88). Higher Body Mass Index was associated with higher odds of underreporting (AOR = 1.09, 95% CI: 1.01–1.18), whereas higher number of reported snacks between meals was associated with lower odds of underreporting (AOR = 0.81, 95% CI: 0.76–0.86). Underreporting increased over the seven-day period (51.4% on day 1 to 65.9% on day 7).

Conclusions

This study highlights a high prevalence of energy intake underreporting among Lao PDR adolescents when using the seven-day estimated food record method. While this prospective method has advantages, its accuracy depends on sustained participant engagement and motivation. Future research is needed to develop, adapt and validate tools, such as technology-assisted tools, for use among adolescents in Lao PDR.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40795-026-01416-y.

Introduction

Adolescence represents a critical stage of growth and development marked by rapid physiological changes and increased nutritional requirements compared with younger school-aged children [1]. Dietary behaviors established during this period often persist into adulthood and influence long-term health outcomes [2]. Accurate dietary assessment is therefore essential for understanding adolescents’ dietary patterns to inform the design and evaluation of interventions that promote optimal nutrition and prevent future diet-related health issues [3].

Self-reported dietary assessment methods such as food-frequency questionnaires (FFQ), 24-hour dietary recalls, and food records are widely used in dietary research due to their practicality in large-scale studies [4]. However, these methods are prone to measurement error. Retrospective approaches, such as FFQ and 24-hour dietary recalls, are particularly vulnerable to recall bias, as participants must rely on memory to report past intake [5]. In contrast, errors in prospective methods, such as food records, are mainly related to misestimation of portion sizes or omission of consumed foods. Underreporting is common, whereas overreporting, although less frequent, can also occur [5, 6]. Adolescents are particularly vulnerable to such errors because of irregular eating habits, snacking, and social desirability bias [7, 8].

Misreporting is not merely a measurement issue but may also affect the interpretation of epidemiological findings. Previous studies have shown that energy intake (EI) misreporting can distort observed relationships between dietary exposures and health outcomes, potentially leading to biased epidemiological inferences and inappropriate analytical decisions [9, 10]. In addition, EI misreporting may reflect broader inaccuracies in the reporting of foods and nutrients, resulting in biased estimates of nutrient inadequacy and dietary risk at the population level [10, 11]. Understanding the prevalence and determinants of EI misreporting is therefore essential for interpreting dietary data and informing appropriate decisions in nutritional epidemiology.

The Goldberg cut-off method is commonly used to identify EI misreporting, particularly when biomarker methods such as doubled-labeled water are not feasible or when objective measures of total energy expenditure and physical activity level (PAL) are unavailable [12, 13]. The Goldberg cut-off method identifies misreporting by comparing reported EI with estimated energy requirements [14]. Previous studies using this method have found a high prevalence of EI underreporting among adolescents (17%–48%) across various dietary assessment tools [7, 15–17]. This variability likely reflects differences in dietary assessment methods, recording periods, participant characteristics, and assumptions used to identify implausible reporters. Consequently, comparisons across studies should be interpreted cautiously, and findings from high-income settings may not be directly applicable to low- and middle-income country (LMIC) populations with different dietary behaviors, food environments, and cultural practices. Understanding the magnitude and determinants of misreporting in specific contexts is therefore crucial for improving the validity of dietary assessment methods and the interpretation of dietary data. Yet the majority of existing studies were conducted in high-income Western settings, with limited data from adolescents in LMICs [18–21].

In Lao People’s Democratic Republic (PDR), a LMIC in Southeast Asia, nutrition research remains limited, and little is known about the validity of dietary assessment methods for this population. This study aims to address this knowledge gap by examining the prevalence and predictors of EI misreporting using a seven-day estimated food record (EFR) dietary assessment tool among adolescents in Lao PDR. By identifying factors associated with the under- and over-reporting, this study contributes new evidence to support the refinement of dietary assessment practices in a LMIC adolescent population.

Methodology

Study design, settings, and participants

This study used secondary data from a cross-sectional survey conducted by the Lao Tropical and Public Health Institute (Lao TPHI) in which the food of students attending eight secondary schools in the Phonhong District of the Vientiane Province, Lao PDR, was assessed from March to April 2019. The original study was designed to assess anemia and the nutritional status of adolescents. The sample size was calculated based on the estimated prevalence of anemia [22]. The eight schools were randomly selected from a total of 37 schools within the Phonhong district, and participants were randomly selected from the student rosters of each school, as detailed previously [22]. Eligible participants were apparently healthy adolescents aged 10–18 years old who were able to communicate in the Lao language. For female participants, eligibility additionally required having reached menarche and not being pregnant at the time of data collection. Participation was voluntary, and written informed consent was obtained from all participants and their caregivers. Of the eligible students approached for participation, 97.8% agreed to participate and completed the survey, resulting in a final sample of 405 students [22].

Measurements

Dietary intake

Food consumption data were assessed using a self-reported EFR collected daily for seven consecutive days. Trained enumerators from Lao TPHI instructed participants to record their food and beverage consumption on the forms they provided, including specifics such as food names, quantities (measured in household utensil units like cups, bowls, spoons, or pieces), cooking methods, and the time and location of consumption. Participants were encouraged to record all meals, snacks, and beverages from their first meal/snack upon waking up until their last meal/snack before going to bed. Daily visits by enumerators to the schools ensured continuous review of the forms completed by each participant. To improve accuracy, data collectors cross-referenced portion sizes of food items with participants using a photographic atlas containing images of food items alongside their corresponding portion sizes estimated in household utensil measurements and grams. The photographic atlas was developed by the National Institute of Public Health based on data collected in the 2016 National Food Consumption Survey. Any discrepancies or unfamiliar descriptions recorded in the forms were promptly addressed and clarified with the respondents. When adolescents had forgotten to record their food intake, data collectors prompted them to recall and record their food and beverage consumption from the previous day.

Using the above collected data, we converted the portion sizes of each food/beverage item, as measured in household utensil units, to grams utilizing the referenced photographic atlas. We further converted portion sizes into their caloric equivalents, using energy values from various food composition tables (FCT) including the Lao FCT, Thailand FCT, Indonesia FCT, and Southeast Asian FCT [23]. For analytical purposes, the foods were grouped into 15 categories, including (1) added fats and sugars, (2) breakfast cereals, (3) mixed dishes, (4) grain products, (5) starchy roots, (6) fruits, (7) vegetables, (8) legumes, nuts, and seeds, (9) herbs, spices, and pastes, (10) meat, fish, and eggs, 11) dairy products, 12) sweet and savory snacks, 13) alcoholic beverages, 14) sugar-sweetened beverages, and 15) other beverages.

Snacking frequency

We summarized the number of morning snacks (between breakfast and lunch), afternoon snacks (between lunch and dinner), and dinner snacks (after dinner) consumed per week from the EFR for each participant.

Anthropometry and demography

We used sociodemographic information such as sex, age, ethnicity, school type, number of household members, and parental occupations, which were collected through interviewer-administered structured questionnaires.

We calculated body mass index (BMI) based on the height and weight collected from each participant by trained enumerators. Weight was recorded with precision to the nearest ± 0.1 kg using a battery-powered digital scale (Seca 874), while height was measured to the nearest ± 0.1 cm using a stable stadiometer designed for mobile height assessment (Seca 213).

Data analysis

Plausible- and implausible-reporting classification

We applied the Goldberg cut-off equation, as detailed by Black [14], to categorize the participants into plausible reporters and implausible reporters (i.e., underreporters and overreporters) of EI. The cut-off limits were calculated using the following equation:

where EI_rep is the reported mean EI (kcal/day) for each individual. BMR is the basal metabolic rate (kcal/day), estimated separately for female adolescents ((13.384 x weight) + 692.6) and male adolescents ((17.686 x weight) + 658.2) at the individual level using the Schofield equation [24, 25].

Because PAL was not directly measured in this study, age- and sex-specific light PAL values (1.45–1.60) were applied following the WHO/FAO/UNU energy requirement guidelines [25] in the primary analysis. These values were consistent with the high prevalence of low physical activity among Lao adolescents (78% of boys and 91% of girls) [23]. To assess the sensitivity of the classification results to the use of light PAL values, we reran the analyses using age- and sex-specific moderate PAL values (1.70–1.85).

SD_min and SD_max were set to -1.96 and + 1.96, respectively to represent the 95% confidence interval, and the value of n was 1 as the data were collected at the individual level.

The S factor accounts for variability in intake, BMR, and energy requirements, where:

• CV_wEI is the pooled mean coefficient of variation of EI within persons (derived from the mean and standard deviation of EI data within this study),

• d is the number of days of dietary recording,

• CV_wB represents within-subject variation in BMR (7.5 for boys and 9.3 for girls), and CV_tP is the total variation in energy requirement (9.5–17.8), based on age-sex-specific values from Black [14].

Participants with EI: BMR ratios below the lower cut-off were classified as underreporters, those within the limits were classified as plausible reporters, and those above the upper limit were classified as overreporters.

In addition, we analyzed the percentage of plausible reporters, underreporters, and overreporters of EI, for each day over the seven-day dietary intake recording period, using this equation. We kept the S value derived from seven-day EI data while adjusting the EI_rep value based on each day’s EI_rep.

Dietary intake

For each participant, average daily EI (kcal/day) from each food group was calculated using all available dietary recording days. Missing recording days were excluded from the analysis. Food groups that were not consumed on a given recording day were assigned a value of 0 kcal. Each food group’s contribution to EI was analyzed for both as energy intake per day (kcal/day) and as a percentage of the total daily EI, which were then averaged across all available recording days for each participant.

Statistical analysis

We analyzed data using Stata SE version 18.0. Statistical analyses followed standard epidemiological procedures. Categorical variables were summarized as frequencies and proportions, while continuous variables were reported as medians and 25th and 75th percentile due to non-normal distributions. Differences between groups were assessed using the Mann-Whitney U-test. Because only one participant was classified as an overreporter in the primary analysis, the overreporter was excluded from analyses examining factors associated with underreporting and the subsequent analyses.

Factors associated with EI underreporting were examined using univariable and multivariable logistic regression. Variables included in the multivariable model were selected a priori based on evidence from their potential relevance to EI misreporting among adolescents, including demographic characteristics (sex, age group, and ethnicity), nutritional status indicator (BMI), socioeconomic and environmental factors (e.g., parental occupation, household size, and school type), and dietary behavior (number of snacks between meals per week). Given the relatively small number of candidate explanatory variables, all selected variables were entered simultaneously into the multivariable model. Adjusted odds ratios (AORs) and 95% confidence intervals (CIs) were reported. Associations between serving size and EI reporting status were further analyzed using multivariable linear regression, adjusting for sex and age group. No missing values were observed for the outcome variable or covariates included in the regression models; therefore, all observations were included in the model. Multicollinearity among independent variables included in the final multivariable model was assessed using variance inflation factors (VIFs), with values > 5 considered indicative of problematic multicollinearity. Model fit was evaluated using the Hosmer–Lemeshow goodness-of-fit test, with p > 0.05 indicating adequate model fit.

For the sensitivity analysis, the prevalences of underreporting, plausible reporting, and overreporting were recalculated using the alternative PAL assumption (i.e., moderate instead of light activity values). Multivariable logistic regression analyses were subsequently repeated using the revised classifications of reporting status to evaluate the robustness of factors associated with underreporting.

To test the trend in underreporting across the 7-day recording period, daily reporting status was recoded into a binary variable: underreporter and non-underreporter (i.e., plausible reporter and overreporter). This binary classification was used because the purpose of the analysis was specifically to test whether the probability of underreporting increased over time. A mixed-effect logistic regression model was fitted with recording day (days 1–7) as a continuous predictor and participant identification number as a random intercept to account for repeated daily observations within individuals. Missing daily dietary records were retained as missing and excluded only from the corresponding day-specific observation, while all available observations from the same participant were included in the analysis.

Results

Participant characteristics

Among the 405 participants, approximately 51% were female and aged 15–18 years old. The majority of participants were from the Lao ethnic group, attended the public schools, resided in households with 6 household members or less, and had parents who were self-employed (Table 1).

Table 1.

Characteristics of the participants by their energy intake reporting status

Variables	Total n (%)	Under reporters n (%)	Plausible reporters n (%)	Over reporters n (%)
Reporters	405 (100.0)	248 (61.2)	156 (38.5)	1 (0.3)
Sex
Male	199 (49.1)	131 (65.8)	68 (34.2)	0 (0.0)
Female	206 (50.9)	117 (56.8)	88 (42.7)	1 (0.5)
Age group (years)
10–14	199 (49.1)	130 (65.3)	68 (34.2)	1 (0.5)
15–18	206 (50.9)	118 (57.3)	88 (42.7)	0 (0.0)
Ethnicity
Lao	314 (77.5)	184 (58.6)	129 (41.1)	1 (0.3)
Minorities	91 (22.5)	64 (70.3)	27 (29.7)	0 (0.0)
School type
Public	300 (74.1)	186 (62.0)	114 (38.0)	0 (0.0)
Private	45 (11.1)	22 (48.9)	22 (48.9)	1 (2.2)
Boarding	60 (14.8)	40 (66.7)	20 (33.3)	0 (0.0)
Household members
5 and below	258 (63.7)	156 (60.5)	101 (39.2)	1 (0.4)
6 and above	147 (36.3)	92 (62.6)	55 (37.4)	0 (0.0)
Mother’s occupation
Self-employed	301 (74.3)	100 (63.1)	110 (36.5)	1 (0.3)
Office worker	51 (12.6)	27 (52.9)	24 (47.1)	0 (0.0)
Housewife	53 (13.1)	31 (58.5)	22 (41.5)	0 (0.0)
Father’s occupation
Self-employed	307 (75.8)	191 (62.2)	115 (37.5)	1 (0.3)
Office worker	98 (24.2)	57 (58.2)	41 (41.8)	0 (0.0)
Body mass index
Median (p25–p75)	19.1 (17.1–21.1)	19.3 (17.1–21.5)	18.9 (17.1–0.6)	14.5 (–)
Reporting snacks between meals (times/week)
Median (p25–p75)	5.0 (2.0–8.0)	2.0 (5.0–9.0)	7.0 (5.0–10.0)	9.0 (–)

Energy intake misreporting

Under the primary analysis using a light PAL assumption, 61.2% of the participants were classified as EI underreporters, whereas only 0.3% (1 participant) were classified as overreporters (Table 1). In the sensitivity analysis using a moderate PAL assumption, the prevalence of underreporting increased from 61.2% to 81.5%, whereas no participants were classified as overreporters (Supplementary Table S1).

Under the primary analysis, trends in EI reporting over seven days showed that underreporting was lowest on the first day (51.4%), gradually increasing to its peak on the sixth day (71.9%). Conversely, the percentage of plausible reporters followed the opposite pattern (Fig. 1). Mixed-effects logistic regression confirmed a significant trend in increasing rates of underreporting over the 7-days of dietary data collection (OR = 1.15, 95% CI: 1.10–1.20, p < 0.001).

Fig. 1.

Percentage of underreporters, plausible reporters, and overreporters on each day over seven days for 405 participants

Factors associated with energy intake underreporting

Factors significantly associated with EI underreporting included age group, body mass index, and snacking frequency (Table 2). Older adolescents (15–18 years old) had a 46% lower odds of underreporting compared with younger adolescents (10–14 years old) (AOR = 0.54; 95% CI: 0.34–0.88). Higher BMI was positively associated with increased odds of underreporting; each one-unit increase in BMI was associated with a 9% increase in the odds of being classified as an underreporter (AOR = 1.09; 95% CI: 1.01–1.18). In contrast, more frequent snacking between meals was associated with lower odds of underreporting, with each additional snacking occasion corresponding to a 19% reduction in the odds of underreporting (AOR = 0.81, 95% CI: 0.76–0.86). No other socio-demographic characteristics were significantly associated with EI underreporting (Table 2). No evidence of problematic multicollinearity was observed among variables included in the final model (all VIFs ≈ 1), and the Hosmer–Lemeshow test indicated adequate model fit (p = 0.348).

Table 2.

Factors associated with underreporting of energy intake

Variables	Crude OR (95% CI)	P-value	Adjusted OR (95% CI)	P-value
Sex
Male	Ref		Ref
Female	0.69 (0.46–1.03)	0.071	0.95 (0.60–1.50)	0.839
Age group (years)
10–14	Ref		Ref
15–18	0.70 (0.46–1.04)	0.084	0.54 (0.34–0.88)	0.014*
Ethnicity
Lao	Ref		Ref
Minorities	1.66 (1.00–2.74)	0.048*	1.19 (0.67–2.12)	0.541
School type
Public	Ref		Ref
Private	0.61 (0.32–1.15)	0.131	0.58 (0.29–1.18)	0.135
Boarding	1.22 (0.68–2.20)	0.495	1.73 (0.88–3.41)	0.107
Household members
5 and below	Ref		Ref
6 and above	1.08 (0.71–1.64)	0.708	1.00 (0.70–1.70)	0.872
Mother’s occupation
Self-employed	Ref		Ref
Office worker	0.65 (0.35–1.18)	0.160	0.56 (0.27–1.18)	0.133
Housewife	0.81 (0.45–1.47)	0.502	0.78 (0.40–1.51)	0.474
Father’s occupation
Self-employed	Ref		Ref
Office worker	0.83 (0.52–1.33)	0.452	0.97 (0.55–1.73)	0.939
Body mass index	1.06 (1.00–1.13)	0.048*	1.09 (1.01–1.18)	0.020*
Number of snacks between meals per week	0.81 (0.77–0.86)	< 0.001*	0.81 (0.76–0.86)	< 0.001*

Overreporters (n = 1) were excluded from this analysis

Univariable and multivariable logistic regression was used to obtain Odds Ratio (OR) and P-values

*p < 0.05

Results of the sensitivity analysis showed that the overall pattern of factors associated with underreporting remained largely unchanged. Age group, BMI, and snacking frequency remained significantly associated with underreporting, although ethnicity additionally reached statistical significance under the moderate PAL assumption (Supplementary Table S2).

Differences in energy intake and portion size by food group

The median EI among all participants was 1196.2 kcal/day, and it was significantly higher among plausible reporters compared with underreporters (i.e., 1584.8 kcal/day vs. 1012.5 kcal/day). Plausible reporters also demonstrated significantly higher median EI (kcal/day) from most food groups compared to underreporters (Table 3).

Table 3.

Comparison of energy intake from each food group between plausible reporters and underreporters

Food groups	Total (n = 404)a	Plausible reporters (n = 156)	Underreporters (n = 248)	P-valueb
	Median (p25–p75)	Median (p25–p75)	Median (p25–p75)
Total energy intakec(kcal)	1196.2 (964.1–1510.1)	1584.8 (1358.8–1927.2)	1012.5 (863.3–1162.5)	< 0.001*
Added fats and sugars (kcal)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.239
Breakfast cereals (kcal)	0.0 (0.0–0.0)	0.0 (0.0–7.8)	0.0 (0.0–0.0)	0.122
Mixed dishes (kcal)	8.6 (3.4–19.5)	10.1 (4.1–22.3)	7.6 (3.0–18.2)	0.065
Grain products (kcal)	545.0 (419.4–752.5)	739.8 (545.5–1060.5)	482.0 (361.2–620.7)	< 0.001*
Starchy roots (kcal)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.189
Fruits (kcal)	24.4 (0.0–54.7)	34.1 (12.2–65.6)	17.9 (0–44.3)	< 0.001*
Vegetable (kcal)	15.7 (6.8–29.8)	17.6 (9.5–34.6)	14.4 (5.4–25.6)	0.001*
Legumes, nuts, and seeds (kcal)	0.0 (0.0–11.7)	0.0 (0.0–20.4)	0.0 (0.0–8.4)	0.004*
Herbs, spices & pastes (kcal)	6.6 (1.3–16.6)	9.4 (2.6–20.3)	5.1 (1.0–14.4)	< 0.001*
Meat, fish, and eggs (kcal)	290.3 (206.9–392.2)	372.5 (274.4–460.0)	247.1 (184.1–335.8)	< 0.001*
Dairy products (kcal)	0.0 (0.0– 9.4)	0.0 (0.0– 7.9)	0.0 (0.0– 0.4)	0.046*
Sweet and Savory snacks (kcal)	19.7 (0.0–73.6)	53.4 (8.0–125.7)	7.5 (0.0–49.3)	< 0.001*
Alcoholic beverages (kcal)	0.0 (0.0–0)	0.0 (0.0–0)	0.0 (0.0–0.0)	0.010*
Sugar-sweetened beverages (kcal)	61.2 (29.7–99.2)	75.2 (43.9–116.5)	53.2 (21.9–86.3)	< 0.001*
Other beverages (blended and non-blended dairy) (kcal)	23.3 (0.0–82.3)	47.6 (0.1–107.7)	0.9 (0.0–65.4)	< 0.001*

*p < 0.05

^a Overreporters (n = 1) were excluded from this analysis

^b Mann-Whitney U test was used to obtain the P-value

^c Mean energy intake of 7 days

Plausible reporters demonstrated a significantly higher %EI from sweet and savory snacks, alcoholic beverages, and other beverages than underreporters. In contrast, underreporters exhibited a significantly higher %EI derived from the meat, fish, and eggs food group than the plausible reporters (Table 4). Plausible reporters reported significantly higher portion sizes of foods from almost all food groups except for sweet and savory snacks, vegetables, and mixed dishes (Supplementary Figure S1).

Table 4.

Comparison of the percentage of Energy Intake (%EI) contributed by each food group between plausible reporters and underreporters

Food groups	Total (n = 404)a	Plausible reporters (n = 156)	Underreporters (n = 248)	P-valueb
	Median (p25–p75)	Median (p25–p75)	Median (p25–p75)
Added fats and sugars (%EI)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.313
Breakfast cereals (%EI)	0.0 (0.0–0.0)	0.0 (0.0–0.4)	0.0 (0.0–0.0)	0.313
Mixed dishes (%EI)	0.8 (0.3–1.8)	0.7 (0.3–1.6)	0.8 (0.3–2.1)	0.056
Grain products (%EI)	48.5 (38.5–57.7)	48.3 (36.3–57.7)	48.5 (39.7–57.7)	0.744
Starchy roots (%EI)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.213
Fruits (%EI)	1.8 (0.0–4.4)	2.1 (0.6–4.0)	1.6 (0.0–4.8)	0.176
Vegetable (%EI)	1.3 (0.6–2.7)	1.2 (0.6–2.3)	1.4 (0.6–3.0)	0.350
Legumes, nuts, and seeds (%EI)	0.0 (0.0–0.9)	0.0 (0.0–0.9)	0.0 (0.0–0.6)	0.036*
Herbs, spices & pastes (%EI)	0.6 (0.1–1.4)	0.7 (0.2–1.4)	0.5 (0.1–1.4)	0.346
Meat, fish, and eggs (%EI)	24.7 (19.3–30.2)	22.6 (18.4–28.8)	25.8 (20.1–31.4)	0.002*
Dairy products (%EI)	0.0 (0.0–1.4)	0.0 (0.0–1.9)	0.0 (0.0–0.8)	0.136
Sweet and Savory snacks (%EI)	1.8 (0.0–6.0)	3.3 (0.4–7.4)	0.7 (0.0–4.2)	< 0.001*
Alcoholic beverages (%EI)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.018*
Sugar-sweetened beverages (%EI)	5.2 (2.4–8.7)	5.0 (2.7–8.3)	5.4 (1.9–8.9)	0.871
Other beverages (blended and non-blended dairy) (%EI)	1.7 (0.0–6.1)	2.9 (0.0–6.6)	0.2 (0.0–5.9)	0.001*

^*p < 0.05

^a Overreporters (n = 1) were excluded from this analysis

^b Mann-Whitney U test was used to obtain the P-value

Discussion

Our study found that 61.2% of adolescents in Lao PDR underreported their EI when assessed using a seven-day EFR. Older adolescents were less likely to underreport than younger adolescents, and those who reported snacking more frequently were also less likely to underreport. In contrast, underreporting tended to be higher among adolescents with higher BMI. The percentage of underreporting was lowest on the first day of dietary assessment and increased progressively over time. Underreporters also reported a lower percentage of EI from snacks and beverages but a higher percentage from meat, fish, and egg compared to plausible reporters, and they generally reported lower total EI across most food groups.

The 61.2% underreporting observed in our study is higher than most estimates reported in adolescent studies using the Goldberg cut-off. Previous studies applying this method show substantial variability, with a French study using a seven-day EFR reporting 26% underreporting [7], Malaysian FFQ-based studies reporting 17.4% [17], and studies among Caucasian children and African American girls using three nonconsecutive day 24-hour dietary recalls reporting 28% and 48%, respectively [15, 16]. Studies using alternative plausibility approaches to identify EI underreporters have also identified high levels of underreporting, including 67% for UK adolescents [18] and 51% for Chilean adolescents [26]. The higher prevalence of EI underreporting observed in our study compared with many previous adolescent studies may reflect the difference in dietary assessment tools, population characteristics, contextual factors, and sources of systematic error. The use of a seven-day EFR may have increased participant burden and reporting fatigue, as reflected in the progressive increase in underreporting across consecutive days in our study. Cultural eating practices may also have contributed to misreporting. Eating from shared dishes or central bowls is common in many LMICs [27], including Lao PDR, which may make it challenging for adolescents to accurately estimate the quantities of food consumed. In addition, many commonly consumed Lao dishes are mixed dishes prepared using multiple ingredients and condiments, which may make it difficult for adolescents to accurately recall and quantify individual components of the meal.

In addition to participant-related challenges, systematic errors in EI estimations may have arisen from inaccurate food composition data. Lao PDR does not have a comprehensive FCT, which meant food composition data were adapted from regional and neighbouring countries. Although efforts were made to improve the accuracy by adjusting nutrient values for cooking methods and recipe ingredients, inaccuracies may remain, particularly for mixed cooked dishes that were derived from raw ingredients. Furthermore, available food composition databases may not fully capture variations in the ingredients of mixed dishes, and the nutrient composition of newly introduced and reformulated commercially processed foods [28]. Similar challenges have been reported in other resource-constrained settings, where the lack of comprehensive national food composition databases necessitates the use of borrowed or adapted food composition data [29]. Such limitations may introduce error into estimates of dietary EI and contribute to misclassification of reporting status. These challenges highlight the need for continued investment in the development and updating of national food composition databases to improve the quality of dietary assessment and nutrition surveillance in LMICs [29, 30].

These findings also have implications for the design of future dietary assessment methods. Although prospective EFR do not rely on memory, their accuracy depends on participants’ motivation and recording consistency and fatigue may reduce attention to detail, especially for snacks and beverages. Technology-assisted dietary assessment tools may help overcome these limitations [31]. Studies in Vietnam and Ghana have shown the feasibility of using AI-assisted image captured mobile applications for dietary assessment, with reasonable validity compared to weighed food records [32, 33]. Evidence from high income contexts suggests that gamified, smartphone-based food logging can enhance motivation and accuracy among adolescents [34, 35]. Future research should explore the adaptation of such tools for LMIC contexts, including the development of locally relevant food composition databases, consideration of cultural eating practices, and adolescent-friendly design features to enhance both usability and data quality.

In our study, older adolescents (15–18 y) were less likely to underreport their EI compared with younger adolescents (10–14 y), suggesting improved accuracy of dietary reporting with age. This result contrasts with previous studies reporting higher levels of underreporting among older than younger age groups [19, 36, 37]. Differences in age categorization and the level of parental involvement may partly explain this discrepancy. For instance, one study found a more underreporting among adolescents aged 12–19 years compared with younger children aged 2–5 years whose intake was reported with parental assistance [19]. Methodological differences may also contribute to the inter-study differences, as a Japanese study using household dietary records with family assistance reported more underreporting among older than younger adolescents despite similar age categorization to our study [37]. In our study, cognitive maturity and familiarity with a wider range of foods may have facilitated more accurate recording among older than younger adolescents, whereas younger adolescents may have had more difficulty completing food records without parental assistance.

Higher BMI was associated with an increased likelihood of EI underreporting in our study, which is consistent with previous dietary surveys of adolescents across diverse settings, regardless of the dietary assessment method used [7, 16, 17, 19, 38]. Collectively, these findings underscore the importance of accounting for weight-related biases when assessing and interpreting dietary intake data of adolescent populations.

This study showed that higher reported snacking frequency between meals was associated with lower odds of underreporting. One possible explanation was underreporters may have omitted recording snacks and beverages, resulting in lower reported snacking frequency. This interpretation is consistent with the lower reported energy contribution, both in absolute energy intake and as proportion of total energy intake, and the higher percentage of energy from main meal ingredients such as meat, fish and eggs among underreporters compared to plausible reporters. Similar trends have been documented among adolescents in other settings, where snacks and beverages are often underreported [18, 26, 39, 40]. Alternatively, the observed association may reflect true differences in snacking behaviors among underreporters and plausible reporters. Given the cross-sectional nature of the analysis and the reliance on self-reported dietary data, the direction of this association and the mechanisms underlying it remain uncertain. Future studies using complementary or more objective dietary assessment approaches may help clarify the mechanisms underlying this finding. Regardless of the underlying mechanism, this finding highlights the potential importance of snacking-related behaviors in understanding dietary reporting accuracy among adolescents in this setting.

The high prevalence of underreporting observed in this study has important implications for future dietary research among Lao adolescents. Given that more than half of participants were classified as underreporters, estimates of dietary intake derived from self-reported food records should be interpreted with caution. Previous studies have shown that the inclusion of implausible reporters can influence observed associations between dietary exposures and health outcomes. For example, one study found that the association between dietary energy density and cardiometabolic risk factors in adolescents became apparent only after excluding implausible reporters [9]. However, analytical approaches to address misreporting are not straightforward. Excluding implausible reporters can introduce selection bias, while adjusting for total EI may lead to differential bias if specific food groups are differentially misreported [41]. Nevertheless, these findings highlight the importance of considering reporting plausibility when analyzing and interpreting dietary data in adolescent populations.

The findings of this study should also be interpreted in the context of the method used to identify implausible energy reporting. We applied the Goldberg cut-off method, which remains one of the most widely used approaches in nutritional epidemiology due to its practicality and low cost, particularly in large population-based studies. Direct measurement is the gold standard; however, it requires specialized equipment and is often impractical in large-scale surveys, particularly in resource-constrained settings. Consequently, predictive equations are widely adopted, for identifying implausible energy intake estimations but they introduce uncertainty in the estimation of energy requirements and reporting status classification. Previous validation studies have demonstrated that the performance of Goldberg cut-offs is influenced by assumptions regarding BMR and PAL, and that some individuals may be misclassified as underreporters or plausible reporters when these equations are used [13, 42, 43]. Consistent with previous methodological studies [13, 42, 43], our sensitivity analysis showed that the estimated prevalence of underreporting increased substantially when a moderate PAL assumption was applied instead of a light PAL assumption, highlighting the sensitivity of prevalence estimates to PAL selection. Nevertheless, the main factors associated with underreporting remained broadly consistent across PAL assumptions, suggesting the identified determinants of underreporting were relatively robust. While the exact prevalence of underreporting should be interpreted cautiously and viewed as an indicator of reporting plausibility at the population level rather than a definitive classification of individual participants [14], the findings consistently indicate that underreporting is common among adolescents in this setting. These findings reinforce the importance of examining the plausibility of energy intake estimates and conducting sensitivity analyses when objective measures of physical activity are unavailable.

From a public health perspective, this issue is particularly relevant in LMICs, where objective approaches such as doubled labelled water, indirect calorimetry, or accelerometer-based estimation of energy expenditure are often prohibitively expensive and logistically challenging for large-scale surveys. Consequently, methods such as the Goldberg cut-off continue to play an important role in assessing data quality and identifying implausible reporting in resource-constrained settings. However, researchers should recognize the uncertainty associated with BMR estimation and PAL assumptions, and conduct a sensitivity analysis where feasible, and then interpret the prevalence estimates of underreporting within this methodological context.

Limitations

This study is the first to investigate the patterns and determinants of EI underreporting among adolescents in Lao PDR. Several limitations should be acknowledged when interpreting the prevalence of underreporting reported in this study. First, this study was conducted among adolescents attending schools in peri-urban areas using a seven-day EFR as the dietary assessment method. Thus, its generalizability to both other dietary assessment methods and to adolescents from different settings is limited. Second, the analysis of factors associated with underreporting was cross-sectional; therefore, causal relationships cannot be established, and the direction of observed associations should be interpreted cautiously. Third, although a range of sociodemographic and behavioral factors were examined, residual confounding from unmeasured variables remains possible.

Recommendations and future research

Future research in similar resource-constrained contexts should apply a short adolescent-specific PAL questionnaire to classify an individual’s PAL for use in the Goldberg cut-off computation. Where feasible, an objective measure, such as accelerometers, could provide more accurate PAL estimates and improve the identification of EI misreporting. Recent studies indicate technology-assisted dietary assessment tools, particularly AI-supported image capture mobile applications, might enhance data quality and adherence among adolescents [32–35]. Future research should assess their feasibility in Lao PDR, incorporating local food databases, to improve dietary assessment accuracy. Expanding research to include adolescents from urban and rural areas, as well as comparing different dietary assessment approaches, would provide more comprehensive insight into the choice of dietary assessment methods for estimating the energy and nutrient intakes of adolescents in Lao PDR.

Conclusion

This study is the first to examine EI misreporting among adolescents in Lao PDR, contributing evidence related to the validity of seven-day EFR for the assessment of dietary intakes in a LMIC context. The high prevalence of EI underreporting (61.2%) exceeds most estimates from previous adolescent studies, using Goldberg cut-offs, yet still falls within the wide variability observed across settings and misreporting identification techniques. As comparable evidence comes primarily from high-income countries, these findings highlight the need for additional dietary assessment validation studies in LMICs where the prevalence of inadequate nutrient intakes is likely high. In our study, older adolescents (15–18 y) were less likely to be classified as underreporters than younger adolescents (10–14 y), while a relatively high BMI was associated with increased odds of underreporting. More frequent snacking was associated with lower odds of underreporting. Underreporters also reported a lower energy contribution from snacks and beverages and smaller portion sizes across most food groups, suggesting they might not capture between-meal consumption and experienced challenges in estimating portion sizes. Increases in underreporting over the seven-day EFR period suggest that participant fatigue and reduced motivation were contributing factors. The findings point to the need for future research to determine whether a shorter than seven-day diet recording period or technology-assisted dietary assessment tools would reduce under-reporting and improve the accuracy of dietary intake estimations in the Lao context. Generating accurate and reliable dietary intake data is essential, for informing policies and interventions, that aim to address the rising rates of overweight and obesity and persistently high prevalence of micronutrient deficiencies among adolescents in Lao PDR.

Abbreviations

BMI

Body mass index

BMR

Basal metabolic rate

EFR

Estimated food records

EI

Energy intake

FCT

Food composition table

FFQ

Food frequency questionnaire

Lao TPHI

Lao Tropical and Public Health Institute

LMIC

Low- and middle-income country

PAL

Physical activity level

Lao PDR

Lao People’s Democratic Republic

%EI

Percentage of energy intake

Authors’ contributions

TK, EF and CT conceptualized the study; TK, MV, and KM analyzed the data; TK and EF interpreted the data; TK drafted the manuscript; EF, MS and CT revised the manuscript. All authors read and approved the final manuscript.

Funding

None.

Data availability

The data that support the findings of this study are available from Lao Tropical and Public Health Institute, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of Lao Tropical and Public Health Institute.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles of the Declaration of Helsinki. The study utilized secondary data collected as part of a separate project conducted by the Lao Tropical and Public Health Institute, which received ethical approval from the National Ethical Committee for Health Research of Lao PDR (Protocol No. 013/2019/NECHR). Informed consent was obtained from all adolescent participants and their legal guardians before the data were collected. The study strictly followed the relevant guidelines and regulations. The dataset is fully anonymized and de-identified, ensuring that no individual participants could be identified or traced.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.